Rescue mode

Rescue mode is a network-boot environment on your VPS when the guest operating system will not start. You mount the disk, fix files such as fstab, reset a root password, or copy data off, then reboot into the guest again. It is a temporary repair workshop, not a factory reset and not a faster form of SSH. Do not run production services from rescue. If the guest still boots, stay in the guest and use rescue only when it will not start.

When VPS rescue is the right move

Use rescue after an fstab UUID typo, a kernel panic on boot, a forgotten root password the console cannot fix, or a full disk that fails fsck without manual help. You may also need files off the disk before a rebuild. The panel boots a small rescue image and attaches your old disk as a second device. You SSH into that environment, mount the disk, make your edits, unmount cleanly, then leave rescue so the original kernel can try again.

Dedicated servers use IPMI and an ISO instead of this button. Shared hosting is handled through a support ticket because you do not boot the node. A password reset in rescue means mounting the root filesystem and running chpasswd or passwd --root against it, then rebooting. That is not a format or a wipe. Rebuild installs a clean template and wipes the disk you came here to fix, so choose carefully.

When rescue is the wrong move

Do not enable rescue only because SSH feels slow or you want a second look. The site goes down because the guest is no longer answering requests. Networking looks different because you are on the rescue NIC. Cron and your normal services are not running because this is not the guest. Some people then start nginx from the rescue image against the mounted disk and lose that work on the next reboot. Finish the repair and leave rescue promptly.

What to fix before you leave

If you changed disks and copied a bad fstab, mount the filesystem in rescue, correct the UUID so it matches a real volume, unmount, and boot again. If the filesystem is damaged beyond repair, use rescue to copy off what remains, then rebuild. If you only needed an earlier state, use a snapshot revert instead of rescue. Write down whether you are in rescue or in the guest before you paste commands into a ticket. The shell prompt can look similar late at night, but the disk layout does not.